Thesis (MSc (Mathematical Sciences))--University of Stellenbosch, 2011. / Includes bibliography. / Please refer to full text to view abstract.

One of the most important parameters of speech is the fundamental frequency of vibration of voiced sounds. The audio sensation of the fundamental frequency is known as the pitch. Depending on the tonal/non-tonal category of language, the fundamental frequency conveys intonation, pragmatics and meaning. In addition the fundamental frequency and intonation carry speaker gender, age, identity, speaking style and emotional state. Accurate estimation of the fundamental frequency is critically important for functioning of speech processing applications such as speech coding, speech recognition, speech synthesis and voice morphing. This thesis makes contributions to the development of accurate pitch estimation research in three distinct ways: (1) an investigation of the impact of the window length on pitch estimation error, (2) an investigation of the use of the higher order moments and (3) an investigation of an analysis-synthesis method for selection of the best pitch value among N proposed candidates. Experimental evaluations show that the length of the speech window has a major impact on the accuracy of pitch estimation. Depending on the similarity criteria and the order of the statistical moment a window length of 37 to 80 ms gives the least error. In order to avoid excessive delay as a consequence of using a longer window, a method is proposed ii where the current short window is concatenated with the previous frames to form a longer signal window for pitch extraction. The use of second order and higher order moments, and the magnitude difference function, as the similarity criteria were explored and compared. A novel method of calculation of moments is introduced where the signal is split, i.e. rectified, into positive and negative valued samples. The moments for the positive and negative parts of the signal are computed separately and combined. The new method of calculation of moments from positive and negative parts and the higher order criteria provide competitive results. A challenging issue in pitch estimation is the determination of the best candidate from N extrema of the similarity criteria. The analysis-synthesis method proposed in this thesis selects the pitch candidate that provides the best reproduction (synthesis) of the harmonic spectrum of the original speech. The synthesis method must be such that the distortion increases with the increasing error in the estimate of the fundamental frequency. To this end a new method of spectral synthesis is proposed using an estimate of the spectral envelop and harmonically spaced asymmetric Gaussian pulses as excitation. The N-best method provides consistent reduction in pitch estimation error. The methods described in this thesis result in a significant improvement in the pitch accuracy and outperform the benchmark YIN method.

The population of bottlenose dolphins off the east coast of Scotland has been studied since the late 1980s, initially focused on the inner Moray Firth, where a Special Area of Conservation (SAC) was designated under the EU Habitats Directive. The population has since expanded its distributional range and currently ranges from the Moray Firth to the Firth of Forth. The main aims of this thesis were: (1) to estimate population parameters for this population using a 25 year individual recognition dataset, and (2) to increase knowledge of the distribution and abundance of dolphins in areas outside the SAC, especially to investigate areas of high use in St Andrews Bay. Apparent survival rate for adults and sub–adult dolphins was estimated at 0.946 (SE=0.005) accounting for temporary emigration caused by the population's range expansion. Sex-specific survival was estimated for males (0.951, SE=0.013) and females (0.956, SE=0.011) using multistate models to minimize bias caused by individuals of unknown sex. Using a newly developed approach, fecundity rate was estimated at 0.222 (95% CI=0.218-0.253) from an expected mean inter-birth interval of 4.49 yrs (95% CI=3.94-4.93). Total population size was estimated as ~200 individuals, after accounting for temporary emigration and for heterogeneity in capture probabilities. In St Andrews Bay, an area used regularly in summer by approximately half the estimated population, habitat use modelling identified the entrance to the Firth of Tay and waters around Montrose as high use areas for dolphins, whose presence was influenced by tidal current speed and direction. The results suggest that the conservation and management plan for this small and isolated population of bottlenose dolphins should be reviewed to adapt it to current knowledge, especially regarding the uncertainty around the potential impacts of offshore renewable energy developments off the east coast of Scotland.

This thesis is about the task of unsupervised parser induction: automatically learning grammars and parsing models from raw text. We endeavor to induce such parsers by observing sequences of terminal symbols. We focus on overcoming the problem of frequent collocation that is a major source of error in grammar induction. For example, since a verb and a determiner tend to co-occur in a verb phrase, the probability of attaching the determiner to the verb is sometimes higher than that of attaching the core noun to the verb, resulting in erroneous attachment *((Verb Det) Noun) instead of (Verb (Det Noun)). Although frequent collocation is the heart of grammar induction, it is precariously capable of distorting the grammar distribution. Natural language grammars follow a Zipfian (power law) distribution, where the frequency of any grammar rule is inversely proportional to its rank in the frequency table. We believe that covering the most frequent grammar rules in grammar induction will have a strong impact on accuracy. We propose an efficient approach to grammar induction guided by cross-linguistic language parameters. Our language parameters consist of 33 parameters of frequent basic word orders, which are easy to be elicited from grammar compendiums or short interviews with naïve language informants. These parameters are designed to capture frequent word orders in the Zipfian distribution of natural language grammars, while the rest of the grammar including exceptions can be automatically induced from unlabeled data. The language parameters shrink the search space of the grammar induction problem by exploiting both word order information and predefined attachment directions. The contribution of this thesis is three-fold. (1) We show that the language parameters are adequately generalizable cross-linguistically, as our grammar induction experiments will be carried out on 14 languages on top of a simple unsupervised grammar induction system. (2) Our specification of language parameters improves the accuracy of unsupervised parsing even when the parser is exposed to much less frequent linguistic phenomena in longer sentences when the accuracy decreases within 10%. (3) We investigate the prevalent factors of errors in grammar induction which will provide room for accuracy improvement. The proposed language parameters efficiently cope with the most frequent grammar rules in natural languages. With only 10 man-hours for preparing syntactic prototypes, it improves the accuracy of directed dependency recovery over the state-ofthe- art Gillenwater et al.’s (2010) completely unsupervised parser in: (1) Chinese by 30.32% (2) Swedish by 28.96% (3) Portuguese by 37.64% (4) Dutch by 15.17% (5) German by 14.21% (6) Spanish by 13.53% (7) Japanese by 13.13% (8) English by 12.41% (9) Czech by 9.16% (10) Slovene by 7.24% (11) Turkish by 6.72% and (12) Bulgarian by 5.96%. It is noted that although the directed dependency accuracies of some languages are below 60%, their TEDEVAL scores are still satisfactory (approximately 80%). This suggests us that our parsed trees are, in fact, closely related to the gold-standard trees despite the discrepancy of annotation schemes. We perform an error analysis of over- and under-generation analysis. We found three prevalent problems that cause errors in the experiments: (1) PP attachment (2) discrepancies of dependency annotation schemes and (3) rich morphology. The methods presented in this thesis were originally presented in Boonkwan and Steedman (2011). The thesis presents a great deal more detail in the design of crosslinguistic language parameters, the algorithm of lexicon inventory construction, experiment results, and error analysis.

A tornado is a violent rotating column of air hanging from the cloud base of a thunderstorm. They occur most frequently in the United States, about 1000 tornadoes every year, but they can also take place in other parts of the world. For a tornado to arise the atmosphere has to be in a specific state. The state includes a steep lapse rate, a significant difference in wind speed between the ground and at 1 km altitude and moisture represented by a low cloud base level. With these three parameters present there is a good chance that a tornado will form. A tornado alone is not possible to predict with the technology that is available today, but the parameters that were just mentioned are. With climate models and Representative Concentration Pathways (RCP-scenarios) the future climate can be investigated based on greenhouse gas emissions among other things. In this research occurrence of the three parameters in the area of the United States with most tornado activity during a period of 30 years in present time was compared to a period of 30 years in the future for two different RCP-scenarios. The result was calculated to see how many times and days the three parameters occurred both at the same time and individually. It was seen that the amount of times they occurred corresponded to what was expected from the theory, about 1000 every year for both time periods. Thereby no difference was seen in the frequency of tornadoes today compared to any of the future scenarios. A difference that could be determined however was the amount of days in which they occurred. The criteria for a tornado was met in 623 days for present time, compared to RCP 8.5, which is the RCP-scenario with the largest amount of greenhouse gas emissions, where the criteria was met 655 days. RCP 2.6 that is the least devastating scenario with the smallest amount of greenhouse gas emissions had the lowest amount of 604 possible tornado days. / En tromb är ett väderfenomen som består av en intensivt roterande luftpelare mellan ett åskmoln och jordytan. De förekommer över hela världen men är vanligast i USA där det bryter ut ungefär 1000 stycken varje år. För att en tromb ska uppstå krävs det att atmosfären i området har några speciella egenskaper. Den ska vara instabilt skiktad, vilket kan förknippas med ett kraftigt temperaturavtagande med höjd, ha stor skillnad i vindhastighet mellan marken och 1 km upp, samt ha hög relativ luftfuktighet vilket indikeras av en låg molnbas. Är dessa tre kriterier uppfyllda samtidigt finns en möjlighet för en tromb att bildas. Enskilda trombers existens kan inte förutspås med dagens teknik eftersom de är kortlivade väderfenomen och inte en del av ett större vädersammanhang. Med dagens klimatmodeller och Representative Concentration Pathways (RCP-scenarier) kan däremot olika väderparametrars förekomst i framtiden undersökas bl.a. med avseende på mängden utsläpp av växthusgaser.      I den här studien har de tre ovan nämnda parametrarnas förekomst analyserats för det mest trombdrabbade området i USA och jämförts under en 30-årsperiod i nutid med en 30-årsperiod i framtiden för två olika RCP-scenarier. Resultatet togs fram utifrån hur många tillfällen och dagar som de tre parametrarna uppfylldes både enskilt och samtidigt. Antalet tillfällen de uppfylldes samtidigt motsvarade statistiken över trombförekomst relativt bra med ungefär 1000 gånger per år i båda tidsperioderna. Det var därmed ingen större skillnad mellan antalet potentiella tromber idag jämfört med något av RCP-scenarierna i framtiden. Däremot var det en viss skillnad på hur många dagar det finns en chans att det bildas tromber. Kriterierna uppfylldes under 623 dagar i nutid jämfört med 655 dagar i RCP-8.5, vilket är scenariot med störst andel utsläpp av växthusgaser. Under RCP-2.6, det mildaste scenariot med minst utsläpp, uppfylldes kriterierna under 604 dagar, färre än både nutid och RCP 8.5.

Determinations of design parameters and investigation on operation performance of a tar removal system for gas cleaning of biomass producer gas have been undertaken. The presence of the tars in the producer gas has been the major hindrance for the commercialisation of the biomass gasification technology for power generation, hydrogen production, Fischer Tropsch (FT) synthesis, chemical synthesis and synthetic natural gas (SNG) synthesis. The characteristic of the tars to condense at reduced temperatures cause problems in the downstream processing as the tars can block and foul the downstream process equipment such as gas engines reactor channels, fuel cells, etc. Considerable efforts have been directed at the removal of tars from the producer gas where the tars can be either chemically converted into lighter molecular weight molecules or physically transferred from gas phase to liquid or solid phase. In the former, the tars have been removed in a scrubber by transferring them from the producer gas to a scrubbing liquid and then removed from the liquid to air in a stripper and finally recycled them into air to a gasifier to recover their energy. A tar removal test system involving a scrubber and stripper has been designed based on the predicted tar solubility in canola methyl ester (CME) as the scrubbing liquid and its measured properties (CME is a type of methyl ester biodiesel). The tar solubility has been predicted to decrease with increasing temperatures and thus its value increases at lower temperatures. In designing the test system, the design parameters are needed including equilibrium coefficients of the gas-liquid system, molar transfer coefficient and the optimum liquid to gas flow rate ratio. The equilibrium coefficients have been predicted based on thermodynamic theories where the required data are determined from CME composition and known properties of each component of the CME as well as the properties of the model tar (naphthalene). The molar transfer coefficients are then experimentally determined and the correlations as a function of liquid and gas flow rates are proposed which are consistent with literature. The optimum liquid to gas flow rate ratios have been found to be 21.4±0.1 for the scrubber and 5.7±0.1 for the stripper. Using these optimum ratios, the tar removal efficiencies in the scrubber and the stripper are 77 and 74%, respectively. The analysis of the system performance has been achieved after an innovative method of determining tar concentrations in both the liquid and gas phase had been developed based on the concept of the density of liquid mixtures. However, these tar removal efficiencies are low due to the fact that the targeted tar concentration in the scrubber’s off-gas was large. As a result the system has been redesigned based on the determined design parameters and its operation performance retested. In the redesigned system, the tar removal efficiency in the scrubber and stripper is 99%. The redesigned system would be integrated with the UC gasifier for downstream gas cleaning. Since 1% of tars are not removed, a makeup tar free CME of 0.0375 litres per hour for the 100kW UC gasifier has been introduced in the recycle stream between the scrubber and stripper to avoid tar accumulation in the system.

No description available.

No description available.

No description available.

The impacts of current forest management methods on surface water quality, especially in northern parts of Sweden, are largely unexplored. In this review reports linked to logging impacts on catchment biogeochemistry, (with special emphasis on boreal ecosystems) has been assessed. Logging disturbances in boreal forest catchments can change biogeochemical processes in soils by alter transpiration, soil conditions, temperature, soil microbial activities and water fluxes. Combined these changes can cause increased soil nutrient leaching to receiving waters. In the studies reviewed, dissolved organic carbon (DOC) concentrations and export generally increased after logging. Similar responses followed for phosphorous (P), but to a lesser extent for nitrogen (N). Streamflow and peakflow and the export of suspended matter (SM) can also be altered after logging. Removal of streamside vegetation may result in increased stream water temperatures and potentially affect lake water temperature, deepen the epilimnion and deplete oxygen (O2) concentrations. Increased wind exposure can resuspend sediments and potentially release bioavailable P. Affected lake water concentrations of N can be considered negligible after logging, whereas tot-P concentrations has shown to increase to an almost twofold level in some studies. The overall impact on the pelagic productivity are therefore most likely those connected to increased DOC and SM concentrations. The general impact on fish biomass can be considered insignificant. In conclusions, to improve future forest management and for further understanding concerning the biogeochemical environmental impacts that forestry might have on freshwaters, additional studies are still required.